description: Disables the mixed precision graph rewrite. (deprecated)

tf.train.experimental.disable_mixed_precision_graph_rewrite

View source on GitHub

Disables the mixed precision graph rewrite. (deprecated)

tf.train.experimental.disable_mixed_precision_graph_rewrite()

Warning: THIS FUNCTION IS DEPRECATED. It will be removed after 2020-11-30. Instructions for updating: Use tf.keras.mixed_precision. There is a guide at https://www.tensorflow.org/guide/mixed_precision. Alternatively, tf.compat.v1.mixed_precision.disable_mixed_precision_graph_rewrite can be used, but this is not recommended for TF2 code.

After this is called, the mixed precision graph rewrite will no longer run for tf.functions, and so float32 operations will no longer be converted to float16.

This does not undo the effects of loss scaling. Any optimizers wrapped with a LossScaleOptimizer will continue to do loss scaling, although this loss scaling will no longer be useful, as the graph rewrite no longer converts tf.functions to use float16.

This function is useful for unit testing. A unit test can test using the mixed precision graph rewrite, then disable it so future unit tests continue using float32.